It isn’t just the forum. Other systems on the servers are slow too. I’m investigating.
I think one of the servers had a serious problem…all three were up and running seemingly OK. I rebooted server3 and nothing changed, but when I rebooted server2 everything seemed to spring back into life. The server also took longer to reboot than it should do so I’m wondernig if there was a disk corruption that needed to be repaired before the O/S could load.
Things seem to be back to normal for now, but I’m still watching!
1 Like
I should add that the forum runs on server1 so server2 must have been affecting the whole server environment. I think that last time there was a problem rebooting one of the servers seemed to solve the problem. Unfortunately I don’t remember whether that was server2. If it was then perhaps that server has a hardware fault. All three servers were obtained through the auction which means they’ve been used before. That usually means they’ve just become surplus to someone’s requirememts (like the two servers I used before we moved to the current three), but it’s possible someone returned one because it was faulty and didn’t notify the hosting company.
You cant blame that on Windows updates… 
1 Like
I think from your previous posts you rebooted all three the last time
1 Like
I recall doing that, but I don’t know what order I did them in. Whilst it seems logical to do 1, 2, 3, I may have used a different order. For example, if I thought server2 was currently running a GFS download then I might have left that until last, or it could just be random based on which servers I have open SSH sessions on. I’ll do the ones that have open sessions first before opening other sessions.
“Maybe we are barking at the wrong tree” Could it be a cloud-flare problem?
This morning I was unable to access the discourse-WW website.
Error code 524 in a partly filled window
Empty window
→ inspect window showed error code 524 for latest.json
Another one with very long wait-times “latest.json” and “poll” over 30 seconds to load
But accessing the site from farther away → VPN USA in California or Washington DC
→ no problems.
Even quitting and reloading or browsing multiple topics
→ no problem with farther away VPN-entry points.
Why should it work in the USA and not via the Brussels one?
It could be a Cloudflare but assuming people connect through the closest Cloudflare location that would mean issues from Brussels, London, Manchester and Edinburgh. It could be a European network issue, but there’s nothing listed on the Cloudflare status pages reporting a widespread issue.
I don’t know how Cloudflare routing works internally, but there are three Cloudflare tunnels to my servers, one to each server. So perhaps Cloudflare in Europe was routing traffic through the server2 tunnel and in the USA it was routing through server3? Server2 seemed to be the one with issues, so that ‘broke’ European connections whilst US connections carried on working? I’m not aware of any diagnostic information which shows me which server a connection goes through.
Rebooting obviously re-links the tunnel on the server so maybe that resolved the problems? If it happens again I will try restarting the tunnel on each server before rebooting to see if that solves the problem.
Are they the Argo tunnels that Cloudflare reported problems with the other day?
Difficult to say. What I’m using used to be called Argo Tunnels, but Argo is now a separate product that I don’t subscribe to. From a Cloudflare branding perspective I’m using a Zero Trust Tunnel, so if they mentioned an Argo Tunnel they may mean something else.
It was the day (7 July) when you were wrestling with “a firewall and QUIC protocol issue”:
I’d never heard of them so I tried Wikipedia and got this: Argo Tunnel - Wikipedia
thats a bit of an unfortunate name
1 Like
I guess they haven’t reworded all the error messages yet. That wasn’t a Cloudflare issue though. The tunnel (Argo/Zero Trust) wasn’t working because the end of it on my servers wasn’t responding/wasn’t responding correctly.
That’s what happens when you let IT people pick names for things. Zero Trust is a fairly big thing in cyber-security at the moment.